This paper describes a new algorithm for the automatic identification of the type of partial discharge sources in insulation systems of rotating machine stator windings. The proposed algorithm is based on fuzzy rules that associate each partial discharge phenomenon to a fuzzy space class trough a tree path selection. The output is a simple frame where the defect type and a membership are evidenced. Applications of the algorithm to analyze data collected both on field and in laboratory are reported and discussed to show its validity and robustness.
